通过机器学习模型和可穿戴传感器数据在模拟团队运动中估计氧气吸收:试点研究
Dermot Sheridan1, Arne Jaspers2,3, Dinh Viet Cuong1,4
1School of Computing, Dublin City University, Dublin, Ireland.
PloS one
|April 21, 2025
概括
机器学习模型可以通过团队运动中的可穿戴传感器来估计氧气吸收 (VO2). 这项技术提供了一种非侵入性的方法来监测运动员的训练状态和生理需求.
科学领域:
- 运动科学 运动科学 运动科学
- 生物技术是生物技术.
- 机器学习 机器学习
背景情况:
- 优化训练状态和降低团队运动中的受伤风险需要精确的生理监测.
- 目前用于评估生理需求的方法在培训期间可能是侵入性的或不切实际的.
研究的目的:
- 研究使用带有可穿戴传感器的机器学习 (ML) 模型来估计团队运动运动员的氧气吸收 (VO2) 的可行性.
- 用不同的传感器数据和功能来比较各种ML模型对VO2估计的性能.
主要方法:
- 六名男子团队运动运动员接受了增量健身测试.
- 收集的数据包括惯性测量单位 (IMU),心率和呼吸率.
- 多个ML模型 (MLR,XGBoost,LSTM,CNN,MLP) 用原始和工程IMU功能进行训练和评估.
主要成果:
- 使用IMU原始数据的长短期记忆 (LSTM) 模型在VO2预测中取得了最高的准确性 (RMSE: 4.976,MAE: 3.698).
- 多重线性回归 (MLR) 模型显示出具有竞争力的性能,特别是在工程特征方面.
- 多传感器设置,包括干和四肢位置,提高了预测准确度.
结论:
- 机器学习模型显示了在团队体育运动中进行非侵入性,实时VO2监测的巨大潜力.
- 穿戴式传感器数据与ML相结合,可以为运动员的生理需求提供宝贵的见解.
- 这种方法可以帮助优化培训并减轻受伤风险.
相关概念视频
Pulse Oximetry
286
Pulse oximetry, or SpO2, is a non-invasive method for continuously monitoring arterial oxygen saturation (SaO2). This procedure involves attaching a probe or sensor to the patient's fingertip, forehead, earlobe, or nose bridge. The sensor works by detecting changes in oxygen saturation levels through light signals generated by the oximeter and reflected by the pulsing blood under the probe.
Purpose
Average SpO2 values are greater than 95%. If the readings fall below 90%, it indicates that...
Purpose
Average SpO2 values are greater than 95%. If the readings fall below 90%, it indicates that...
286
Exercise and Cardiac Output
702
Regular physical activity is essential for maintaining cardiovascular health, with aerobic exercises being particularly effective. According to the American Heart Association, 150 minutes of moderate to intense aerobic exercise per week is recommended for a healthy heart. Aerobic activities may include brisk walking, running, bicycling, cross-country skiing, and swimming, ideally performed three to five times per week.
Sustained exercise increases the muscles' oxygen demand, which can be...
Sustained exercise increases the muscles' oxygen demand, which can be...
702


